Use temporary subclasses to test Interpolated distribution

Author: a-rodin

pymc3/tests/test_distributions.py

@@ -793,14 +793,29 @@ def test_multidimensional_beta_construction(self):
         with Model():
             Beta('beta', alpha=1., beta=1., shape=(10, 20))
 
-    @pytest.mark.skip('Interpolated logp testing is not implemented yet')
     def test_interpolated(self):
         for mu in R.vals:
             for sd in Rplus.vals:
-                with Model() as model:
-                    x_points = np.linspace(mu - 5 * sd, mu + 5 * sd, 100)
-                    pdf_points = sp.norm.pdf(x_points, loc=mu, scale=sd)
-                    dist = Interpolated('dist', x_points=x_points, pdf_points=pdf_points)
-                    # TODO evalute logp at x_points somehow
-                    # pdf_output_points = <evaluated logp>
-                    # assert_almost_equal(pdf_output_points, pdf_points, decimal=3)
+                #pylint: disable=cell-var-from-loop
+                xmin = mu - 5 * sd
+                xmax = mu + 5 * sd
+
+                class TestedInterpolated (Interpolated):
+
+                    def __init__(self, **kwargs):
+                        x_points = np.linspace(xmin, xmax, 100000)
+                        pdf_points = sp.norm.pdf(x_points, loc=mu, scale=sd)
+                        super(TestedInterpolated, self).__init__(
+                            x_points=x_points,
+                            pdf_points=pdf_points,
+                            **kwargs
+                        )
+
+                def ref_pdf(value):
+                    return np.where(
+                        np.logical_and(value >= xmin, value <= xmax),
+                        sp.norm.logpdf(value, mu, sd),
+                        -np.inf * np.ones(value.shape)
+                    )
+
+                self.pymc3_matches_scipy(TestedInterpolated, R, {}, ref_pdf)

pymc3/tests/test_distributions_random.py

@@ -578,27 +578,24 @@ def ref_rand(size, mu, beta):
         pymc3_random(pm.Gumbel, {'mu': R, 'beta': Rplus}, ref_rand=ref_rand)
 
     def test_interpolated(self):
-        alpha = 0.05
-        size = 10000
-        tries = 10
-
-        # Interpolated model doesn't support variables as inputs, so it is necessary
-        # to use custom code instead of pymc3_random for it
         for mu in R.vals:
             for sd in Rplus.vals:
-                with pm.Model():
-                    x_points = np.linspace(mu - 5 * sd, mu + 5 * sd, 100)
-                    pdf_points = st.norm.pdf(x_points, loc=mu, scale=sd)
-                    dist = pm.Interpolated('dist', x_points=x_points, pdf_points=pdf_points)
-
-                    p = alpha
-                    f = tries
-                    while p <= alpha and f > 0:
-                        s0 = dist.random(size=size)
-                        s1 = st.norm.rvs(size=size, loc=mu, scale=sd)
-                        _, p = st.ks_2samp(s0, s1)
-                        f -= 1
-                    assert p > alpha, 'Failed KS test for mu = %s, sd = %s' % (mu, sd)
+                #pylint: disable=cell-var-from-loop
+                def ref_rand(size):
+                    return st.norm.rvs(loc=mu, scale=sd, size=size)
+
+                class TestedInterpolated (pm.Interpolated):
+
+                    def __init__(self, **kwargs):
+                        x_points = np.linspace(mu - 5 * sd, mu + 5 * sd, 100)
+                        pdf_points = st.norm.pdf(x_points, loc=mu, scale=sd)
+                        super(TestedInterpolated, self).__init__(
+                            x_points=x_points,
+                            pdf_points=pdf_points,
+                            **kwargs
+                        )
+
+                pymc3_random(TestedInterpolated, {}, ref_rand=ref_rand)
 
     @pytest.mark.skip('Wishart random sampling not implemented.\n'
                       'See https://github.com/pymc-devs/pymc3/issues/538')